New Approaches in Reliability Analysis of Complex Systemssociallab.fer.hr/wordpress/wp-content/uploads/2017/... · • V. Levashenko, E. Zaitseva, M. Kvassay, and T. M. Deserno, “Reliability

New Approaches in Reliability Analysis of Complex Systems

Miroslav [email protected]

University of Žilina

Faculty of Management Science and Informatics

Slovakia

18 May 2017, Valencia

Some Intro

May 2017 INNOSOC VALENCIA 2017 WORKSHOP 2/95

Slovakia and Žilina Region

May 2017 INNOSOC VALENCIA 2017 WORKSHOP

Žilina is a regional city and has near 85 thousands inhabitants.

3/95

University of Žilina

• Technical university

• Established in 1953

• About 9000 students and 1500 employees

• More than 70,000 graduates

• Main area of research –transportation

• 7 faculties:• Faculty of Management Science

and Informatics


Faculty of Management Science and Informatics• Study programs:

• informatics, computer engineering, management

• Established in 1990

• About 1500 students and 140 employees

• More than 3500 graduates

• Main area of research – optimization of (transport) networks, decision support systems, biomedicine

• 7 departments:• Department of Informatics - around 15

academics and research fellows who form research community in Computer Science.


Our Team


aour Projects:

▪ FP7-ICT-2013-10. Regional Anesthesia Simulator & Assistant (RASimAs), Reg. no.610425, 2013-2016

▪ Support Systems for Medical Decision Making, Grant of Research & Development Agency (APVV), Reg. no. SK-PL-0023-12, Slovakia-Poland, 2013-2014

▪ Workshop on Biomedical Technologies, Grant of Visehrad Fund V4, 2014

▪ Intelligent Assistance Systems: Multisensor Processing and Reliability Analysis, NATO Collaborative Linkage Grant, Reg.. no. CBP.EAP.CLG 984, 2011-2012

▪ TEMPUS. Advanced Training and Life Long Learning Program in Applied Health Sciences, Reg. No. 543889-TEMPUS-1-2013-1-SE, 2013-2016

▪ TEMPUS. Green Computing and Communications (GreenCo), Reg.No.530270-TEMPUS-1-2012-1-UK, 2012-2015

etc.

Data Mining

o Decision Making Support Systems

o Fuzzy Decision Trees

o Clustering and Classification

Reliability Engineering

o Reliability Analysis

o Importance Measures

o Sensitivity and TestabilityApplication in • medicine

• decision systems

6/95

Our Cooperation


▪ Institute of Biomedical Informatics, University of Information Technology and Management in Rzeszów, Poland(Dr. Krzysztof Pancerz)

▪ VŠB - Technical university of Ostrava, Czech Republic (Prof. Radim Briš, CSc.)

▪ University Medical Centre Utrecht Image Sciences Institute, The Netherlands(Prof. Max A. Viergever)

▪ Aachen University of Technology, Department of Medical Informatics, Germany (Prof. Thomas M. Deserno )

▪United Institute of Informatics Problems, Belarus (Prof. Alexander Tuzikov)

▪ Siberian State Medical University, Russia (Prof. Sergey Karas)

▪ Bay Zoltán Nonprofit Ltd., Hungary, (Dr. Balint Uzsoki )

▪ University of Ioannina, Greece (Dr. Iosif Androulidakis)

▪ Università Campus Bio-Medico di Roma, Italy (Prof. Paolo Soda)

7/95

Safety of Healthcare


Reliability Analysis


SystemReliability indices

and measures• System availability• MTTF, MTBF…• Reliability function• Importance measures• …

System improving

Mathematical model

9/95

Safety of Healthcare

• Medical error is one of the leading causes of death in the US.

• About 8 – 12% of patients admitted to hospital suffer from adverse events whilst receiving healthcare in the EU.


• M. A. Makary and M. Daniel, “Medical error—the third leading cause of death in the US,” BMJ, vol. 353, p. i2139, May 2016• https://ec.europa.eu/health/patient_safety/policy_en

10/95

Safe vs Unsafe


• R. Amalberti, Y. Auroy, D. Berwick, and P. Barach, “Five system barriers to achieving ultrasafe health care,” Annals of Internal Medicine, vol. 142, no. 9, p. 756, May 2005

11/95

Healthcare System

• Composed of many heterogeneous components.

• Problems of data collecting:

• heterogeneous

• uncertain (expert evaluation)

• incompletely specified


• E. Zaitseva, “Reliability analysis methods for healthcare system,” in Human System Interactions (HSI), 2010 3rd Conference on, 2010, pp. 211–216.

12/95

Reliability of Healthcare System

• Mainly qualitative approaches – focus on identification of steps that result in medical error.

• We try to develop a method for quantitative analysis. The method is a combination of tools of:

• reliability analysis,

• logic algebra,

• data mining.


Reliability of Healthcare System


FDT induction

Selection of all

possible cases

• V. Levashenko, E. Zaitseva, M. Kvassay, and T. M. Deserno, “Reliability estimation of healthcare systems using Fuzzy Decision Trees,” in 2016 Federated Conference on Computer Science and Information Systems (FedCSIS), 2016, pp. 331–340.

14/95





• Why systems fail?

• How to develop reliable systems?

• How to measure and test reliability in design, operation and management?

• How to maintain systems reliable, by maintenance, fault diagnosis and prognosis?

⇓

• How to model the system?

• How to quantify system reliability?

• How to represent, model and quantify uncertainties in system behavior?• E. Zio, “Reliability engineering: Old problems and new challenges,” Reliability Engineering & System Safety, vol. 94, no. 2, pp.

125–141, Feb. 2009.

16/95

Reliability – Basic Concepts


t = 0

t

Reliability

Availability

Maintainability

Security

Safety

Maintenance

• Reliability – the probability that the system operates without failure in the interval <0, t>, given that it worked at time 0.

Dependability

17/95



• Maintainability – the probability that the system will be repaired at time t, given that it failed at time 0.

• Maintenance – all actions that allows repairing system (corrective) or preventing its failure (preventive).

t = 0

t

Reliability

Availability

Maintainability

Security

Safety

Maintenance

Dependability

18/95



t = 0

t

Reliability

Availability

Maintainability

Security

Safety

Maintenance

• Availability – the probability that the system is functioning at time t.

Dependability

19/95

Relationship between Reliability, Maintainability and Availability


Availability – complex characteristic, whose computation can be quite complicated, therefore:• average interval availability:

𝐴avg 𝑡 =1

𝑡න0

𝑡

𝐴 𝜏 𝑑𝜏; 𝑡 > 0

• average (steady-state) availability:

𝐴avg = lim𝑡→∞

𝐴avg 𝑡 = lim𝑡→∞

𝐴 𝑡 = 𝐴

• M. Rausand and A. Høyland, System Reliability Theory, 2nd ed. Hoboken, NJ: John Wiley & Sons, Inc., 2004.

20/95



t = 0

t

Reliability

Availability

Maintainability

Security

Safety

Maintenance

• Safety – the probability that the system will either perform its function correctly or will discontinue its operation in a safe way.

Dependability

21/95



t = 0

t

Reliability

Availability

Maintainability

Security

Safety

Maintenance

• Security – the probability that the system is able to resist internal or external threats.

Dependability

22/95



t = 0

t

Reliability

Availability

Maintainability

Security

Safety

Maintenance

• Dependability – the ability of the system to deliver its intended level of service to its users.

Dependability

23/95

Cause and Effect Relationship• Error – a deviation from correctness or accuracy.

• Defect – the departure of a quality characteristic from its specified value that results in a product not satisfying its normal usage requirements.

• Fault – a physical defect, imperfection or flaw that occurs in hardware or software.

• Failure – a non-performance of some action that is due or expected.


• A. Birolini, Reliability Engineering, 5th ed. Springer, 2007.

24/95

Reliability as Complex Problem

The main goal of reliability analysis is to increase the dependability/reliability of a system.


Availability

Dependability

Reliability

Maintainability

Fault-t

ole

rant

Defe

cts

, hazard

,

erro

rs, fa

ult, fa

ilure

Safety

Security

• A. Birolini, Reliability Engineering, 5th ed. Springer, 2007.

25/95

Different Views on Reliability

• Components create system.

• System is served by personnel.

• System and personnel interact with environment.


Two Approaches

• Qualitative – aims to identify, classify and rank the failure modes, or event combinations that would lead to system failures

• Quantitative – aims to evaluate in terms of probabilities the attributes of dependability (reliability, availability, safety)


Methods of Qualitative Analysis

• Checklist

• Preliminary hazard analysis

• Failure Mode and Effect Analysis (FMEA)

• Fault trees

• …


Preliminary Hazard Analysis

• Hazard – a situation with the potential for injury or fatality whereas failure is the actual event, be it hazardous or otherwise. The term major hazard is different only in degree and refers to certain large-scale potential incidents.

• Preliminary hazard analysis is a semi-quantitative analysis that is performed to:

1. identify all potential hazards and accidental events that may lead to an accident;

2. rank the identified accidental events according to their severity;

3. identify required hazard controls and follow-up actions.


• The risk is established as a combination of a given event/consequence and a severity of the same event/consequence. This will enable a ranking of the events/consequences in a risk matrix:

• E. Zio, An Introduction to the Basic of Reliability and Risk Analysis. London, UK: World Scientific, 2007.

29/95

Failure Mode and Effect Analysis (FMEA)• FMEA is a systematic procedure for identifying the modes of

failures and for evaluating their consequences. It is a tabular procedure which considers hazards in terms of single-event chains and their consequences.

• It is a qualitative method, of inductive nature, which aims at identifying those failure modes of the components which could disable system operation or become initiators of accidents with significant external consequences.

• The basic questions which must be answered by the analyst are:

• How can each component or subsystem fail? (What is the failure mode?)

• What cause might produce this failure? (What is the failure mechanism?)

• What are the effects of each failure if it does occur?



30/95

Failure Mode and Effect Analysis

• Once the FMEA is completed, it assists the analyst in:• selecting, during initial stages, various design alternatives with high

reliability and high safety potential;• ensuring that all possible failure modes, and their effects on

operational success of the system, have been taken into account;• identifying potential failures and the magnitude of their effects on

the system;• developing testing and checkout methods.



31/95

Failure Mode and Effect Analysis –Domestic Hot Water System



32/95

Fault Trees• Fault trees represent hierarchical

approach.

• They are useful for both qualitative and quantitative analyses because they:

• force the analyst to actively seek out failure events (success events) in a deductive manner;

• provide a visual display of how the system can fail, and thus aid understanding of the system by persons other than the designer;

• point out critical aspects of systems failure (system success);

• provide a systematic basis for quantitative analysis of reliability.

• The analysis based on fault trees is performed by identification so-called minimal cut sets.


• B. S. Dhillon, Human Reliability and Error in Medicine. Singapore, SG: World Scientific, 2003.• E. Zio, An Introduction to the Basic of Reliability and Risk Analysis. London, UK: World Scientific, 2007.

33/95

Quantitative Analysis

• Principal steps for reliability estimation of complex systems:

1. definition of number of performance levels for the system model;

2. mathematical representation of the system model;

3. quantification of the system model (calculation of indices and measures, for example importance measures);

4. measuring behavior of the system.


Number of Performance Levels


Binary- and Multi-State Systems


Number of Performance Levels

Principal problems for MSS application:

• High Dimension of the MSS :

• Elaboration of new algorithms, methods and indices.


Properties BSS MSS

Exactness +

Computational complexity +

Elaboration +

n

i

im1

37/95

Mathematical Representation


Markov & semi-Markov

model

Monte-Carlo simulation

model

Universal generating

function

Structure function

Structure function based methods for reliability analysis:

• Fault Trees analysis

• Reliability Block Diagram analysis

• Minimal Cut/Path set based methods

38/95

Estimation of Common Measures


Non-repairable systems Repairable systems

Reliability Availability

Failure rate Repair rate

Mean time to failure Mean time to repair Mean time between failures

items ofnumber total

detected is which faults ofnumber 1)( tR

N

i

it

NMTTF

1

1

0 in time items workingofnumber

items failed ofnumber )(

tt

0 in time items failure ofnumber

items restored ofnumber )(

tt

N

i

itN

MTTR1

1

items ofnumber

items workingofnumber )( tA

MTTRMTTFMTBF


39/95

Example


Time L(t) R(t) F(t) λ(t)

0 1 023 102 1.00000 0.00000 0.02258

1 1 000 000 0.97742 0.02258 0.00577

2 994 230 0.97178 0.02822 0.00414

3 990 114 0.96776 0.03224 0.00338

4 986 767 0.96449 0.03551 0.00299

5 983 817 0.96160 0.03840 0.01221

10 971 804 0.94986 0.05014 0.00981

15 962 270 0.94054 0.05946 0.01121

20 951 483 0.93000 0.07000 0.01291

25 939 197 0.91799 0.08201 0.01553

30 924 609 0.90373 0.09627 0.01953

35 906 554 0.88608 0.11392 0.02560

40 883 342 0.86340 0.13660 0.03485

45 852 554 0.83330 0.16670 0.04886

50 810 900 0.79259 0.20741 0.06993

55 754 191 0.73716 0.26284 0.10133

60 677 771 0.66247 0.33753 0.14738

65 577 882 0.56483 0.43517 0.21342

70 454 548 0.44428 0.55572 0.30484

75 315 982 0.30885 0.69115 0.42476

80 181 765 0.17766 0.82234 0.56966

85 78 221 0.07645 0.92355 0.72415

90 21 577 0.02109 0.97891 0.86045

Measures for item in time t = 75:• R(t) = 0.3088• F(t) = 0.6912• λ(t) = 0.4248Mean time to failure:• MTTF = 62.7373

40/95

Boolean Functions and Binary-State Systems


Binary-State System

• n – number of system components

• for i = 1,2,…, n:• xi – state of component i

• 0 – component is failed• 1 – component is functioning

• pi – probability that the i-th component is working• qi – probability of failure of the i-th component


System

xnx1 …… xixixi


42/95

Structure Function


(x1, x2, …, xn) = (x): {0, 1}n {0, 1}

Structure function Boolean function• state of the system at a fixed time• state of component i at the fixed time

• function value

• value of the i-th variable of the function

Tools of Boolean algebra can be used in reliability analysis of binary-state systems.

• Structure function defines system topology:

43/95

Positive and Negative Logic

• Positive logic• Question: When is system functioning?

• Methods: Reliability block diagram, Minimal path sets

• Variable: xi

• Parameters: Survival probabilities, Availability


• Negative logic• Question: When does system fail?

• Methods: Fault tree, Minimal cut sets

• Variable: ¬xi

• Parameters: Failure probabilities, Unavailability

44/95

Representation of Structure Function


(x) = OR(x1, x2)

x1 x2 (x)

0 0 0

0 1 1

1 0 1

1 1 1

0 x1

x2

1

10

0

1

Analytical Description (formula)

Truth Table Binary Decision Diagram

What are pros and cons of these approaches?

45/95

Structure Function – Analytical Description• Logical representation:

(x) = OR(x1, x2)

• Arithmetical representation:

(x) = x1+x2 x1x2

• Logical-probabilistic representation:

• A(p) = p1+p2 p1p2


Structure Function – Analytical Description• Transform the following logical functions into

arithmetical functions:• f1(x) = NOT(x1)

• f2(x) = AND(x1, x2)

• f3(x) = OR(x1, x2, x3)

• f4(x) = OR(x1, AND(x2, x3))

• f5(x) = NOT(XOR(x1, AND(x2, x3)))


Structure Function – Binary Decision Diagram


x1 x2 x3 (x)

0 0 0 0

0 0 1 0

0 1 0 0

0 1 1 0

1 0 0 0

1 0 1 1

1 1 0 1

1 1 1 1

Truth Table Binary Decision Diagram

48/95

Structure Function – Computations based on Binary Decision Diagram


𝐴 = 𝑝1 𝑞2𝑝3 + 𝑝2 𝑈 = 𝑞1 + 𝑝1𝑞2𝑞3

• Try to express the following function in the form of BDD:• f(x) = OR(AND(x1,x3), AND(x2,x4))

49/95

Structure Function – Problems of Binary Decision Diagrams


Good Bad

• Wikipedia: Binary Decision Diagram (https://en.wikipedia.org/wiki/Binary_decision_diagram)

50/95

Series

Reliability Block Diagrams and Typical StructuresParallel


x1 x2 (x)

0 0 0

0 1 1

1 0 1

1 1 1

x1 x2 (x)

0 0 0

0 1 0

1 0 0

1 1 1

Bridge k-out-of-n

x1 x2 x3 (x)

0 0 0 0

0 0 1 0

0 1 0 0

1 0 0 0

1 0 1 1

1 1 0 1

1 1 1 1

How is the structure function

defined?

How does RBD look like?

51/95

Reliability Block Diagrams


(x) = ?A(p) =?

x1

x3

x4

x2

x5 x6

x3

x4

x5 x6

x1 x2

52/95

Fault Trees (again)


a b c d e f

OR OR

OR

OR(OR(a,b,c), OR(d,e,f))

• How do we obtain availability and unavailability?

• Can we transform the fault tree into reliability block diagram?

• B. S. Dhillon, Human Reliability and Error in Medicine. Singapore, SG: World Scientific, 2003.

53/95

Reliability Block Diagram and Fault Tree


• Positive logic• Question: When is system functioning?

• Negative logic• Question: When does system fail?

x1

x3

x2U x1 ( x2 x3 )

x1 x2 x3

U x1 (x2 x3 )

54/95

Minimal Cut Set


• Reliability block diagram:

• A minimal set of components whose failure results in system failure.

• Fault tree:

• A minimal set of events whose occurrence causes occurrence of the top event.

• Identification of the minimal cut sets is equivalent to finding all the prime implicates of the structure function.

• The dual concept is known as Minimal Path Set. The minimal path sets of the system agree with the prime implicants of the function.


55/95

Minimal Cut Sets


x1

x3

x2

x1 x2 x3 (x)

0 0 0 0

0 0 1 0

0 1 0 0

0 1 1 0

1 0 0 0

1 0 1 1

1 1 0 1

1 1 1 1

x1 x2 x3 (x) RBD Cut Minimal cut

0 0 0 0 {c1c2c3} no

0 0 1 0 {c1c2} no

0 1 0 0 {c1c3} no

0 1 1 0 {c1} yes

1 0 0 0 {c2c3} yes

x2

x3

x1

x2

x3

x1

x2

x3

x1

x2

x3

x1

x2

x3

x1

56/95

Importance Analysis


Structural importance

Birnbaum’s importanceCriticality importance

Component criticality

Minimal cut/ path sets

Qualitative Quantitative

Fussell-Vesely’s importance

𝜙(𝑥1, 𝑥2, 𝑥3) 𝑥3

𝑥1 𝑥2 0 1

0 0 0 0

0 1 0 1

1 0 0 1

1 1 0 1

ComponentComponent state

0 1

1 0.5 0.5

2 0.4 0.6

3 0.1 0.9

• W. Kuo and X. Zhu, Importance Measures in Reliability, Risk, and Optimization: Principles and Applications. Chichester, UK: Wiley, 2012.

57/95

Coherent Systems

• Coherent system:• the structure function is non-decreasing in all its

arguments


Series-parallel systems 𝒌-out-of-𝒏 systems

• E. Zaitseva, M. Kvassay, V. Levashenko, and J. Kostolny, “Importance analysis of k-out-of-n multi-state systems based on direct partial logic derivatives,” in ICTERI 2016, 2016, pp. 441–457.

58/95

Noncoherent Systems


• Noncoherent system:• the structure function is non-decreasing in all its

arguments

𝒌-to-𝒍-out-of-𝒏 systems Logic circuits

• M. Kvassay, E. Zaitseva, V. Levashenko, and J. Kostolny, “Reliability analysis of multiple-outputs logic circuits based on structure function approach,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 36, no. 3, pp. 1–1, Mar. 2016..

59/95

Identification of Critical States –Logical Differential Calculus• Classic partial derivative (real field):

𝜕𝑓

𝜕𝑥𝑖𝑥𝑖 = lim

෦𝑥𝑖→𝑥𝑖

𝑓 𝑥𝑖 , 𝒙 − 𝑓 𝑥𝑖 , 𝒙

𝑥𝑖 − 𝑥𝑖

• Boolean partial derivative (GF(2) field):𝜕𝑓

𝜕𝑥𝑖𝑥𝑖 = 𝑓 ഥ𝑥𝑖 , 𝒙 ⨁𝑓 𝑥𝑖 , 𝒙 = 𝑓 1𝑖 , 𝒙 ⨁𝑓 0𝑖 , 𝒙

• Let us prove it.


Logical Differential Calculus

• Some formulae:𝜕 ҧ𝑓

𝜕𝑥𝑖=

𝜕𝑓

𝜕𝑥𝑖𝜕𝑐

𝜕𝑥𝑖= 0

𝜕 𝑓⨁𝑔

𝜕𝑥𝑖=

𝜕𝑓

𝜕𝑥𝑖⨁

𝜕𝑔

𝜕𝑥𝑖𝜕 𝑓⋀𝑔

𝜕𝑥𝑖= 𝑔

𝜕𝑓

𝜕𝑥𝑖⨁𝑓

𝜕𝑔

𝜕𝑥𝑖⨁

𝜕𝑓

𝜕𝑥𝑖

𝜕𝑔

𝜕𝑥𝑖𝜕 𝑓⋁𝑔

𝜕𝑥𝑖= ҧ𝑔

𝜕𝑓

𝜕𝑥𝑖⨁ ҧ𝑓

𝜕𝑔

𝜕𝑥𝑖⨁

𝜕𝑓

𝜕𝑥𝑖

𝜕𝑔

𝜕𝑥𝑖


• S. N. Yanushkevich, D. M. Miller, V. P. Shmerko, and R. S. Stankovic, Decision Diagram Techniques for Micro- and NanoelectronicDesign Handbook, vol. 2. Boca Raton, FL: CRC Press, 2005.

61/95

Boolean Derivatives


• Partial logic derivative:𝜕𝑓 𝒙

𝜕𝑥𝑖= 𝑓 𝑥𝑖 ⨁𝑓 ഥ𝑥𝑖 = 𝑓 𝑥𝑖 𝑓 ഥ𝑥𝑖 ∨ 𝑓 𝑥𝑖 𝑓 ഥ𝑥𝑖

• Direct partial logic derivative:𝜕𝑓 1 → 0

𝜕𝑥𝑖 1 → 0=

𝜕𝑓 0 → 1

𝜕𝑥𝑖 0 → 1= 𝑓 𝑥𝑖 𝑓 ഥ𝑥𝑖

• Inverse partial logic derivative:𝜕𝑓 1 → 0

𝜕𝑥𝑖 0 → 1=

𝜕𝑓 0 → 1

𝜕𝑥𝑖 1 → 0= 𝑓 𝑥𝑖 𝑓 ഥ𝑥𝑖

• S. N. Yanushkevich, D. M. Miller, V. P. Shmerko, and R. S. Stankovic, Decision Diagram Techniques for Micro- and NanoelectronicDesign Handbook, vol. 2. Boca Raton, FL: CRC Press, 2005.

62/95

Direct Partial Boolean Derivatives


𝑓 𝒙 𝜕𝑓 0 → 1 𝜕𝑥1 0 → 1

𝑓 0,0,0 = 0

𝑓 0,0,1 = 0 𝑓 0,1,0 = 0 𝑓 0,1,1 = 1 𝑓 1,0,0 = 1

𝑓 1,0,1 = 0 𝑓 1,1,0 = 1 𝑓 1,1,1 = 1

𝑓 0,0,0 = 0

𝑓 0,0,1 = 0 𝑓 0,1,0 = 0 𝑓 0,1,1 = 1 𝑓 1,0,0 = 1

𝑓 1,0,1 = 0 𝑓 1,1,0 = 1 𝑓 1,1,1 = 1

1

0

1

0

𝑓 𝒙 𝜕𝑓 1 → 0 𝜕𝑥1 1 → 0

1

0

1

0

~

~

~

~

~

~

~

~

𝑓 𝒙 𝜕𝑓 1 → 0 𝜕𝑥1 0 → 1

𝑓 0,0,0 = 0

𝑓 0,0,1 = 0 𝑓 0,1,0 = 0 𝑓 0,1,1 = 1 𝑓 1,0,0 = 1

𝑓 1,0,1 = 0 𝑓 1,1,0 = 1 𝑓 1,1,1 = 1

𝑓 0,0,0 = 0

𝑓 0,0,1 = 0 𝑓 0,1,0 = 0 𝑓 0,1,1 = 1 𝑓 1,0,0 = 1

𝑓 1,0,1 = 0 𝑓 1,1,0 = 1 𝑓 1,1,1 = 1

0

0

0

0

𝑓 𝒙 𝜕𝑓 0 → 1 𝜕𝑥1 1 → 0

0

0

0

0

~

~

~

~

~

~

~

~

63/95

Logical Differential Calculus in Reliability Analysis


𝜕𝜙 𝑗 → ҧ𝑗

𝜕𝑥𝑖 𝑠 → ҧ𝑠= ቊ

1 if 𝜙 𝑠𝑖 , 𝒙 = 𝑗 and 𝜙 ҧ𝑠𝑖 , 𝒙 = ҧ𝑗0 otherwise

Coherent systems Noncoherent systems𝜕𝜙 1→0

𝜕𝑥𝑖 1→0, 𝜕𝜙 0→1

𝜕𝑥𝑖 0→1

𝜕𝜙 1→0

𝜕𝑥𝑖 1→0, 𝜕𝜙 0→1

𝜕𝑥𝑖 0→1, 𝜕𝜙 1→0

𝜕𝑥𝑖 0→1,𝜕𝜙 0→1

𝜕𝑥𝑖 1→0

64/95

Logical Differential Calculus in Reliability Analysis


Coherent systems Noncoherent systems

𝜙 𝒙 = 𝑥1𝑥3⋁𝑥2𝑥3𝜙 𝒙 = 𝑥1 𝑥2⋁𝑥3

Component

Critical

state

vectors

Critical path

vectors

Critical cut

vectors

3 (1,0,.) (1,0,1) (1,0,0)

ComponentCritical state

vectors

State vectors at

which failure of

component is critical

for system failure

State vectors at

which repair of


for system failure

3 (0,1,.) (1,0,.) (0,1,1) (1,0,0)

𝜕𝜙 1 → 0

𝜕𝑥3 1 → 0=

𝜕𝜙 0 → 1

𝜕𝑥3 0 → 1= 𝑥1𝑥2

𝜕𝜙 1 → 0

𝜕𝑥3 1 → 0=

𝜕𝜙 0 → 1

𝜕𝑥3 0 → 1= 𝑥1𝑥2

𝜕𝜙 1 → 0

𝜕𝑥3 0 → 1=

𝜕𝜙 0 → 1

𝜕𝑥3 1 → 0= 𝑥1𝑥2

𝜕𝜙 1 → 0

𝜕𝑥3 0 → 1=

𝜕𝜙 0 → 1

𝜕𝑥3 1 → 0= 0

65/95

Critical State Vectors



𝜙 𝒙 = 𝑥1𝑥3⋁𝑥2𝑥3

𝜙 𝒙 = 𝑥1 𝑥2⋁𝑥3

Component

Critical

state

vectors

Critical

path

vectors

Critical

cut

vectors

1

(.,0,1)

(.,1,0)

(.,1,1)

(1,0,1)

(1,1,0)

(1,1,1)

(0,0,1)

(0,1,0)

(0,1,1)

2 (1,.,0) (1,1,0) (1,0,0)

3 (1,0,.) (1,0,1) (1,0,0)

ComponentCritical state

vectors

State vectors at

which failure of


for system failure

State vectors at

which repair of


for system failure

1 (.,0,0) (.,1,0) (1,0,0) (1,1,0)

2 (0,.,1) (1,.,1) (0,1,1) (1,1,1)

3 (0,1,.) (1,0,.) (0,1,1) (1,0,0)

• M. Kvassay, E. Zaitseva, and V. Levashenko, “Reliability analysis of noncoherent systems based on logical differential calculus,” in Risk, reliability and safety : innovating theory and practice - Proceedings of the European Safety and Reliability Conference, ESREL 2016, 2017, pp. 1367–1374.

66/95

Importance Measures



𝜙 𝒙 = 𝑥1𝑥3⋁𝑥2𝑥3𝜙 𝒙 = 𝑥1 𝑥2⋁𝑥3

SI3↓ = SI3↓

↓ = TD𝜕𝜙 1 → 0

𝜕𝑥3 1 → 0= TD 𝑥1𝑥2

= 0.25

BI3↓ = BI3↓

↓ = Pr𝜕𝜙 1 → 0

𝜕𝑥3 1 → 0= 1 = 𝑝1𝑞2

SI3↓ = SI3↓

↓ + SI3↑↓ = TD

𝜕𝜙 1 → 0

𝜕𝑥3 1 → 0+ TD

𝜕𝜙 1 → 0

𝜕𝑥3 0 → 1

= TD 𝑥1𝑥2 + TD 𝑥1𝑥2 = 0.25 + 0.25 = 0.5

BI3↓ = BI3↓

↓ + BI3↑↓ = Pr

𝜕𝜙 1 → 0

𝜕𝑥3 1 → 0= 1 + Pr

𝜕𝜙 1 → 0

𝜕𝑥3 0 → 1= 1

= 𝑞1𝑝2 + 𝑝1𝑞2

ComponentStructural

importanceBirnbaum’s importance

1 0.75 𝑝2𝑞3 + 𝑝3𝑞2 + 𝑝2𝑝3

2 0.25 𝑝1𝑞3

3 0.25 𝑝1𝑞2

ComponentStructural

importanceBirnbaum’s importance

1 0.5 𝑞3

2 0.5 𝑝3

3 0.5 𝑝1𝑞2+𝑞1𝑝2

67/95

Multiple-Valued Logic Functions and Multi-State Systems


Some Motivations Nonstandard Logics


Is it filled?

Yes (1) No (0)

Yes (2) It may be (1) No (0)

Binary Logic

Fuzzy Logic

MVL

full

partly

empty

69/95

Multi-State System


Components statesperfect working

working

breakdown

System failure

Example of Multi-State System of

two components

and with three states of reliability for

the system and its every components

Perfect system working

or …

System working

or …

70/95

Multi-State System

• n – number of system components

• m – number of system states

• for i = 1,2,…, n:• mi – number of states of component i• xi – state of component i

• 0 – component is failed• mi - 1 – component is perfectly functioning

• pi,s – probability that the i-th component is in state s


System

xnx1 …… xixixi

• A. Lisnianski, I. Frenkel, and Y. Ding, Multi-state System Reliability Analysis and Optimization for Engineers and IndustrialManagers. London, UK: Springer-Verlag London Ltd., 2010.

71/95

Structure Function


(x1, x2, …, xn) = (x): {0, ..., m1-1}... {0, ..., mn-1} {0, ..., m-1}m1 = m2 = … = mn = m homogeneous system

Structure function Logic function• state of the system at a fixed time• state of component i at the fixed time

• function value

• value of the i-th variable of the function

Tools of multiple-valued logic can be used in reliability analysis of multi-state systems

• Structure function defines system topology:

72/95

Different Interpretation of Multi-State Systems


Truth table of structure function

x1 x2 (x) 0 0 0 0 1 1 0 2 1 1 0 1 1 1 2 1 2 2 2 0 1 2 1 2 2 2 2

The structure function MDD

0

x2 x2

1 2 0

0 1 1 2 2

0 x1 2

1

0 x1

x2 x2

1 2

1 2 0

0 0 1 1

2 2

Truth table of structure function

x1 x2 (x) 0 0 0 0 1 1 0 2 2 1 0 1 1 1 1 1 2 2 2 0 2 2 1 2 2 2 2

The structure function MDD

73/95

Basic Characteristics

• Component states probabilities:𝑝𝑖,𝑠 = Pr 𝑥𝑖 = 𝑠 , 𝑖 ∈ 1,2, … , 𝑛 , 𝑠 ∈ {0,1, … ,𝑚𝑖 − 1}

• System state probability:Pr 𝜙 𝒙 = 𝑗 , 𝑗 ∈ {0,1, … ,𝑚 − 1}

• System availability/unavailability:𝐴≥𝑗 = Pr 𝜙 𝒙 ≥ 𝑗

𝑈≥𝑗 = Pr 𝜙 𝒙 < 𝑗, 𝑗 ∈ {1,2, … ,𝑚 − 1}

• Performance utility function:

𝑂 = σ𝑗=0𝑚−1 𝑜𝑗Pr 𝜙 𝒙 = 𝑗 ,𝑜𝑗 – utility attached to state 𝑗


Importance Analysis• Influence of:

• given component state on given system state / availability level• given component on given system state / availability level• given component state on the whole system• given component on the whole system


Structural importance

Birnbaum’s importanceCriticality importance

Component criticality

Minimal cut/ path sets

Qualitative Quantitative

Fussell-Vesely’s importance

• M. Kvassay, E. Zaitseva, and V. Levashenko, “Importance analysis of multi-state systems based on tools of logical differentialcalculus,” Reliability Engineering & System Safety, vol. 165, no. December 2016, pp. 302–316, Sep. 2017.

75/95

Logical Differential Calculus


𝜕𝜙 𝑗 → ℎ

𝜕𝑥𝑖 𝑠 → 𝑟= ቊ

1 if 𝜙 𝑠𝑖 , 𝒙 = 𝑗 and 𝜙 𝑟𝑖 , 𝒙 = ℎ0 otherwise

76/95

Direct Partial Logic Derivatives


𝜕𝜙 𝑗 ↘


1 if 𝜙 𝑠𝑖 , 𝒙 = 𝑗 and 𝑟𝑖 , 𝒙 < 𝑗0 otherwise

𝜕𝜙 ↘


1 if 𝜙 𝑠𝑖 , 𝒙 > 𝑟𝑖 , 𝒙0 otherwise

𝜕𝜙 ℎ≥𝑗 → ℎ<𝑗


1 if 𝜙 𝑠𝑖 , 𝒙 ≥ 𝑗 and 𝑟𝑖 , 𝒙 < 𝑗0 otherwise

Integrated Direct Partial Logic Derivatives:

Type I:

Type II:

Type III:

𝜕𝜙 𝑗 → ℎ


1 if 𝜙 𝑠𝑖 , 𝒙 = 𝑗 and 𝜙 𝑟𝑖 , 𝒙 = ℎ0 otherwise


77/95

Relations between DPLDs



78/95

Example of Structural Importance


Component 𝟑Component state

Average

0 1 2 3

Syst

em

sta

te

0 – – – – –

1 – 0.50 0 0 0.1666

2 – 0.25 0 0.50 0.2500

3 – 0 0.25 0 0.0833

Sum – 0.75 0.25 0.50 0.50

Component 𝟏Component state

Average

0 1

Syst

em

sta

te

0 – – –

1 – 0.25 0.25

2 – 0.25 0.25

3 – 0.25 0.25

Sum – 0.75 0.75

𝜙(𝑥1, 𝑥2, 𝑥3) 𝑥3

𝑥1 𝑥2 0 1 2 3

0 0 0 0 0 0

0 1 0 1 1 2

1 0 0 1 1 2

1 1 0 2 3 3

• M. Kvassay, E. Zaitseva, J. Kostolny, and V. Levashenko, “Importance analysis of multi-state systems based on integrated directpartial logic derivatives,” in 2015 International Conference on Information and Digital Technologies (IDT), 2015, pp. 183–195.

79/95

Data Mining


Knowledge Discovery Process

1. Understanding the problem domain

2. Understanding the data

3. Preparation of the data

4. Data mining

5. Evaluation of the discovered knowledge

6. Using the discovered knowledge

Knowledge discovery is not a linear process.


• K. J. Cios and G. W. Moore, “Medical data mining and knowledge discovery: Overview of key issues,” in Medical Data Mining and Knowledge Discovery, K. J. Cios, Ed. New York, NY: Physica Verlag Heidelberg, 2001, pp. 1–20.

81/95

Understanding the Data


• What data is available?

• Which data will be used?

• Which additional information will be needed?

create a targetdata set

rows (records)

columns(attributes)

82/95

How Much Data?


Linguistic Data

• data mining – sophisticated process with inaccurate data

• linguistic data transforms inaccuracy into vague

• linguistic data is simpler for understanding (models are smaller and simpler)


Experience

Nurse Doctor

1 1

1 3 2 1 2 2

4 1 5 4

Age (numerical)

Age (linguistic)

8 young 20 young 25 adult 40 adult 50 adult 75 old

84/95

Multiple-Valued and Fuzzy Data


1.0

adult

young is 0.7

adult is 0.3

FUZZY

years

0.5

0

1

2

young old

years

MVL

16 20 50 59

25 50

Numeric

A* Linguistic attribute A* (Age)

А*,1(young) А*,2 (adult) А*,3 (old)

x1 = 8 1.0 0.0 0.0

x2 = 20 0.7 0.3 0.0

x3 = 25 0.5 0.5 0.0

x4 = 40 0.0 1.0 0.0

x5 = 50 0.0 0.5 0.5

x6 = 75 0.0 0.0 1.0

85/95

Data Mining


Data Mining

Discovery

Description

Clustering Summarization Visualization

Prediction

Classification

Neural Networks

Bayesian Networks

Decision TreesSupport Vector

MachinesInstance Based

Regression

Verification

Hypothesis testing

86/95

Decision Trees


Tumor

Age

no

non confirmed

confirmed

elder younger

high

low

Heredity

no

high

low

yes low

H(B) - describes the uncertainty of attribute BH(B|Ai) - describes the uncertainty of attribute B when the attribute Ai is givenI(B ; Ai) - is used as to measure the dependence of the attribute B on the attribute Ai and vice-versa

H(B|Ai1)

I(B; Ai1)

H(B)

H(Ai1)

H(B) = I(B; A) + H(B | A)

87/95

Application of Data Mining in Construction of Structure Function


Laparoscopic Surgery Procedure

• System• 0 – non-operational (fatal medical error), • 1 – partially operational (some imperfection), • 2 – fully operational (surgery without any complication).

• Device (m1 = 2): • 0 – failure, and • 1 –functioning.

• Work of anesthesiologist (m2 = 2): • 0 – non-operational (medical error), • 1 – fully operational (without any complication).

• Work of surgeon and the nurse (m3 = m4 = 3), i.e.: • 0 – (the fatal error), • 1 – (sufficient), and • 2 – (perfect or the work without any complication).



89/95

Collection of Data in the Repository



90/95

Dataset and Structure Function



91/95

Construction of Fuzzy Decision Tree



92/95

Structure Function Generation





94/95

This document has been prepared for the European Commission however it reflects the views only of the authors, and the

Commission cannot be held responsible for any use which may be made of the information contained therein.


Documents

New Approaches in Reliability Analysis of Complex Systemssociallab.fer.hr/wordpress/wp-content/uploads/2017/... · • V. Levashenko, E. Zaitseva, M. Kvassay, and T. M. Deserno, “Reliability